mod = 'aa' # copy from dev repo; crop_size=(256x384), S=36
mod = 'ab' # allow more OOB, by updating threshs to 64; export at 384,512; output 256; export as long as we have N//2
mod = 'ac' # N=128
mod = 'ad' # put more info into name; also print rtime
mod = 'ae' # allow trajs to go behind camera during S
Hi
Thank you so much for sharing your great work!
I'm trying to train a similar model on PointOdyssey dataset.
train.tar.gz,train.tar.gz.partaaandtrain.tar.gz.partXXand so on?export_mp4_dataset.pyfor each ofmodand its corresponding training chunk in this code: